A processing method and apparatus for predicting binding sites of molecular fragments

Abstract

The embodiment of the present application relates to a kind of processing method and device for predicting the binding site of molecular fragment, the method comprises: constructing binding site prediction model;Corresponding first data set is recorded as model training data set, and the binding site prediction model is trained based on the first data set;After model training ends, two three-dimensional molecular fragment structures input by user are received as corresponding fragment structure A, B;And the binding site information of fragment structure A, B is identified using binding site prediction model and the identification information is fed back to the current user.By the present application, the prediction efficiency can be improved, and the discovery ability of unknown connection mode can be enhanced.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a processing method and apparatus for predicting the binding sites of molecular fragments. Background Technology

[0002] In fields such as molecular docking, synthetic chemistry, and molecular editing, different binding sites between molecular fragments (or synthetic building blocks) can lead to variations in the binding efficiency and overall molecular properties of the molecules (composed of two molecular fragments or two synthetic building blocks). Therefore, in molecular design tasks such as molecular docking / synthesis / editing, a predictive method is needed to predict the optimal binding sites for two molecular fragments (or two synthetic building blocks). Currently, this prediction method mainly relies on researchers' experience. Specifically, before prediction, quantum computing tools are used to calculate some empirical parameters of the two molecular fragments (or two synthetic building blocks), and then docking site analysis is performed based on the calculation results and a series of pre-set docking rules. Clearly, this conventional prediction method has significant drawbacks: 1) low prediction efficiency; 2) limited ability to discover new connection modes due to the constraints of human experience / rules. Summary of the Invention

[0003] The purpose of this invention is to address the shortcomings of existing technologies by providing a method, apparatus, electronic device, and computer-readable storage medium for predicting binding sites of molecular fragments. This invention constructs a binding site prediction model that performs binding site prediction processing on first and second molecular fragments as model input and outputs corresponding binding probability prediction vectors. A first dataset is constructed through large-scale data collection to train the binding site prediction model. After model training, the binding site prediction model is used to identify the binding site information of any two three-dimensional molecular fragment structures input by the user and the identified information is fed back to the current user. The end-to-end binding site prediction model provided by this invention can improve prediction efficiency and, through large-scale data training, enhance the model's generalization ability and its ability to discover unknown connectivity patterns.

[0004] To achieve the above objectives, a first aspect of the present invention provides a method for predicting the binding sites of molecular fragments, the method comprising:

[0005] A binding site prediction model is constructed; the binding site prediction model is used to perform binding site prediction processing on the first and second molecular fragments input to the model and output the corresponding binding probability prediction vector.

[0006] The model training dataset is constructed and denoted as the corresponding first dataset; and the binding site prediction model is trained based on the first dataset;

[0007] After the model training is completed, the system receives two three-dimensional molecular fragment structures input by the user as corresponding fragment structures A and B; and uses the binding site prediction model to identify the binding site information of fragment structures A and B and feeds back the identification information to the current user.

[0008] Preferably, the first and second molecular fragments are each a three-dimensional molecular fragment structure;

[0009] The first molecular fragment includes a first set of atoms and a first set of chemical bonds;

[0010] The first set of atoms includes multiple first atoms; the atomic attributes of the first atom include at least a first atom identifier, a first atom type, a first atom coordinates, and a first atom chiral attribute set; the first atom identifier is the unique identifier of the current first atom; the first atom type is the chemical element type of the current first atom; the first atom coordinates are the three-dimensional spatial coordinates of the current first atom; the first atom chiral attribute set consists of multiple first chiral attributes; each first chiral attribute consists of a corresponding attribute type and attribute value; if the first atom is a chiral center atom, then all attribute values ​​in the corresponding first atom chiral attribute set are not empty; if the first atom is not a chiral center atom, then all attribute values ​​in the corresponding first atom chiral attribute set are empty;

[0011] The first set of chemical bonds includes multiple first chemical bonds; the chemical bond attributes of the first chemical bond include at least a first chemical bond identifier, a first chemical bond type, and a first bonding atom group; the first chemical bond identifier is a unique identifier for the current first chemical bond; the first chemical bond type includes multiple chemical bond types; the first bonding atom group is composed of the first atom identifiers of the two first atoms connected by the current first chemical bond;

[0012] The second molecular fragment includes a second set of atoms and a second set of chemical bonds;

[0013] The second atom set includes multiple second atoms; the atomic attributes of the second atom include at least a second atom identifier, a second atom type, second atom coordinates, and a second atom chiral attribute set; the second atom identifier is the unique identifier of the current second atom; the second atom type is the chemical element type of the current second atom; the second atom coordinates are the three-dimensional spatial coordinates of the current second atom; the second atom chiral attribute set consists of multiple second chiral attributes; each second chiral attribute consists of a corresponding attribute type and attribute value; if the second atom is a chiral center atom, then all attribute values ​​in the corresponding second atom chiral attribute set are not empty; if the second atom is not a chiral center atom, then all attribute values ​​in the corresponding second atom chiral attribute set are empty;

[0014] The second set of chemical bonds includes multiple second chemical bonds; the chemical bond attributes of the second chemical bond include at least a second chemical bond identifier, a second chemical bond type, and a second bonding atom group; the second chemical bond identifier is a unique identifier for the current second chemical bond; the second chemical bond type includes multiple chemical bond types; the second bonding atom group is composed of the second atom identifiers of the two second atoms connected by the current second chemical bond.

[0015] Preferably, the first and second model input terminals of the binding site prediction model are used to receive the first and second molecular fragments input to the model, respectively, and the model output terminal is used to output the corresponding binding probability prediction vector.

[0016] The binding site prediction model includes an encoding initialization module, a graph neural coding network, and an MLP network;

[0017] The first and second input terminals of the encoding initialization module are connected to the first and second model input terminals, respectively, and the output terminal is connected to the input terminal of the graph neural network; the output terminal of the graph neural network is connected to the input terminal of the MLP network; the output terminal of the MLP network is connected to the model output terminal.

[0018] The encoding initialization module is used to construct a molecular graph G based on the first and second molecular fragments input from the model; and to perform state encoding initialization processing based on the molecular graph G to obtain a corresponding state encoding tensor H0, which is then sent to the graph neural coding network.

[0019] The molecular graph G includes a node set V and an edge set E;

[0020] The node set V consists of N1+N2 nodes v i Composition; the edge set E consists of multiple edges e i,jComposition; 1 ≤ node index i ≤ N1 + N2, 1 ≤ node index j ≤ N1 + N2, j ≠ i, N1 and N2 are the total number of atoms in the first and second molecular segments, respectively; the node v i Corresponding to one of the first atoms or one of the second atoms; the edge e i,j Corresponding to one of the first chemical bonds or one of the chemical bonds;

[0021] The state encoding tensor H0 includes N1+N2 node state vectors h 0,i The node state vector h 0,i With the node v i One-to-one correspondence; the node state vector h 0,i It consists of a set of atomic type features, atomic coordinate features, chiral feature sets, associated chemical bond feature sets, and cross-segment distance feature sets;

[0022] The atom type feature is the current node state vector h. 0,i The one-hot encoding of the first or second atom type corresponding to the first or second atom;

[0023] The atomic coordinate feature is the current node state vector h. 0,i The corresponding atomic coordinates of the first or second atom;

[0024] The chiral feature set consists of multiple chiral attribute features; each chiral attribute feature is the current node state vector h. 0,i The feature encoding vector of one of the first or second chiral attributes in the set of chiral attributes of the first or second atom corresponding to the first or second atom;

[0025] The set of associated chemical bond features consists of one or more associated chemical bond features; the associated chemical bond features include chemical bond type features and chemical bond orientation features; each associated chemical bond feature is associated with the current node state vector h. 0,i The first or second atom corresponds to a first or second chemical bond; the chemical bond type feature is the unique thermal encoding of the first or second chemical bond type of the currently corresponding first or second chemical bond; the chemical bond direction feature is a three-dimensional displacement vector from the currently corresponding first or second atom to another first or second atom indicated in the first or second bonding atom group of the currently corresponding first or second chemical bond.

[0026] The cross-segment distance feature set consists of multiple cross-segment distance features; if the current node state vector h 0,iFor each of the first atoms, each cross-segment distance feature is the square of the straight-line distance between the currently corresponding first atom and a second atom in the second molecular segment; if the current node state vector h 0,i For each second atom, each of the transmolecular distance features is the square of the straight-line distance between the currently corresponding second atom and a first atom in the first molecule segment;

[0027] The graph neural network is composed of M GNN layers connected sequentially, where the total number of coding layers M is a preset positive integer greater than 2; the m-th GNN layer is used to encode the state tensor H input to this layer. m-1 Perform state transition processing to obtain the corresponding state encoding tensor H m And when the current coding layer index m is less than M, the state coding tensor H is... m Send to the (m+1)th GNN layer, and when the current coding layer index m equals M, send the state coding tensor H. m＝M Send to the MLP network; 1 ≤ coding layer index m ≤ M; the state coding tensor H m Includes N1+N2 node state vectors h m,i The node state vector h m,i With the node v i One-to-one correspondence;

[0028]

[0029] Feature d i,j The setting rule is: if node v i v j If the two corresponding atoms belong to the same molecular segment, then the corresponding characteristic d i,j Fixed to 0; if node v i v j If the two corresponding atoms belong to two different molecular segments, then the corresponding characteristic d i,j Fixed to the corresponding cross-segment distance feature;

[0030] f m Let θ be the state transition function of the m-th layer of the GNN. m f is the m-th state transition function m The corresponding model parameters of the m-th GNN layer;

[0031] The MLP network is used to encode the state tensor H based on the output of the Mth layer of the GNN layer. M The combination probability prediction vector is obtained by performing combination probability prediction; the combination probability prediction vector is obtained by predicting the probability p of N1+N2 nodes. iComposition; the node prediction probability p i With the node v i One-to-one correspondence;

[0032] P = f MLP (H M ;θ MLP );

[0033] P is the combined probability prediction vector, f MLP Let θ be the network mapping function of the MLP network. MLP These are the model parameters of the MLP network.

[0034] Preferably, the first dataset includes multiple first data records; the first data record includes a first training molecular fragment, a second training molecular fragment, and a first label vector;

[0035] The first and second training molecular fragments are each a three-dimensional molecular fragment structure;

[0036] The first label vector includes multiple first node labels; the total number of first node labels is the sum of the total number of atoms in the first and second training molecular segments; the label value of the first node label is 1 or 0; only two first node labels in the first label vector have a label value of 1, and the rest are 0, and the two atoms corresponding to the two first node labels with a label value of 1 are not in the same training molecular segment.

[0037] Preferably, the model training dataset is denoted as the corresponding first dataset, which specifically includes:

[0038] Step 51: Collect data on the three-dimensional molecular structures of various compound molecules through multiple publicly available data channels to obtain multiple corresponding first molecular structures;

[0039] Among them, the various types of publicly available data channels include at least various publicly available datasets that publish molecular three-dimensional structures, journals / magazines / scientific and technological literature / papers;

[0040] Step 52: Each of the first molecular structures is taken as the corresponding current molecular structure; molecular fragment sampling is performed on the current molecular structure to obtain multiple corresponding sampling fragments; a round of traversal is performed on all the sampling fragments of the current molecular structure; during this round of traversal, the currently traversed sampling fragment is taken as the corresponding current fragment; other sampling fragments in the current molecular structure that are connected to the current fragment in a single-atom connection mode are extracted to form a corresponding associated fragment set; the current fragment is taken as a corresponding first training molecular fragment, and any sampling fragment in the current associated fragment set is taken as a corresponding second training molecular fragment, and the first training molecular fragment corresponding to the current fragment and each of the second training molecular fragments are combined to form a corresponding first fragment group; at the end of this round of traversal, all the obtained first fragment groups are deduplicated.

[0041] Step 53: Perform a traversal of all the first fragment groups; during this traversal, the currently traversed first fragment group is taken as the corresponding current fragment group; calculate the total number of atoms of the first and second training molecular fragments in the current fragment group to obtain the corresponding current atomic sum; set a first label vector with a vector length matching the current atomic sum; initialize the label values ​​of all first node labels in the current first label vector to 0; after the vector initialization is completed, reset the label values ​​of the two first node labels corresponding to the two fragment connection atoms of the first and second training molecular fragments in the current fragment group to 1; and form a corresponding first data record by the first training molecular fragment, the second training molecular fragment, and the first label vector corresponding to the current fragment group; and at the end of this traversal, form the corresponding first dataset by all the obtained first data records.

[0042] Preferably, training the binding site prediction model based on the first dataset specifically includes:

[0043] Step 61: Based on a preset first segmentation ratio, the first dataset is divided into two sub-datasets, denoted as the first training set and the first evaluation set.

[0044] Wherein, both the first training set and the first evaluation set are composed of multiple first data records; the ratio of the total number of records in the first training set and the first evaluation set satisfies the first segmentation ratio;

[0045] Step 62: Extract the first data record of the first training set as the corresponding current training record;

[0046] Step 63: Input the first and second training molecular fragments of the current training record as the corresponding first and second molecular fragments into the binding site prediction model for binding site prediction processing, and use the binding probability prediction vector output by the model as the corresponding first prediction vector.

[0047] Step 64: Input the first predicted vector and the first label vector of the current training record into a preset first model loss function to calculate the corresponding first loss value;

[0048] Wherein, the loss function of the first model includes at least the binary cross-entropy loss function;

[0049] Step 65: Identify whether the first loss value meets the preset first loss value range; if the first loss value meets the first loss value range, proceed to step 66; if the first loss value does not meet the first loss value range, modulate the model parameters of the graph neural coding network and the MLP network of the binding site prediction model in one round based on the preset first model optimizer in the direction of minimizing the first model loss function, and return to step 63 when the modulation ends.

[0050] Step 66: Identify whether the current training record is the last first data record in the first training set; if yes, proceed to step 67; if no, extract the next first data record in the first training set as the new current training record and return to step 63.

[0051] Step 67: Perform a round of traversal on all the first data records in the first evaluation set; during this round of traversal, take the currently traversed first data record as the corresponding current evaluation record; input the first and second training molecular fragments of the current evaluation record as the corresponding first and second molecular fragments into the binding site prediction model for binding site prediction processing, and take the binding probability prediction vector output by the model as the corresponding second prediction vector; and form a corresponding prediction-label data pair by the second prediction vector and the first label vector of the current training record; and at the end of this round of traversal, calculate the accuracy, precision, recall and F1 score based on all the obtained prediction-label data pairs to obtain the corresponding first accuracy, first precision, first recall and first F1 score;

[0052] Step 68: Identify the first accuracy, first precision, first recall, and first F1 score; if the first accuracy does not meet the preset first accuracy range, or the first precision does not meet the preset first precision range, or the first recall does not meet the preset first recall range, or the first F1 score does not meet the preset first F1 score range, then return to step 62 to continue training; if the first accuracy meets the first accuracy range, and the first precision meets the first precision range, and the first recall meets the first recall range, and the first F1 score meets the first F1 score range, then stop training and confirm that the model training has ended.

[0053] Preferably, the step of using the binding site prediction model to identify the binding site information of fragment structures A and B and feeding back the identification information to the current user specifically includes:

[0054] Step 71: Input the fragment structures A and B as the corresponding first and second molecular fragments into the binding site prediction model for binding site prediction processing, and use the binding probability prediction vector output by the model as the corresponding AB prediction vector.

[0055] Step 72: Calculate the predicted probabilities p of all nodes in the AB prediction vector corresponding to all atoms of the fragment structure A. i Extract them to form the corresponding first probability set; and then extract all the node prediction probabilities p in the AB prediction vector corresponding to all atoms of the fragment structure B. i Extract them to form the corresponding second probability set;

[0056] Step 73, predict the probability p of the node with the largest probability value in the first probability set. i Let p be the corresponding first maximum probability; and let the node with the largest probability value in the second probability set be predicted as p. i The corresponding second maximum probability is recorded; and the atom identifier in the fragment structure A corresponding to the first maximum probability is used as the corresponding fragment A-side binding atom identifier; and the atom identifier in the fragment structure B corresponding to the second maximum probability is used as the corresponding fragment B-side binding atom identifier;

[0057] Step 74: The A fragment identification information is composed of the atomic identifier and the corresponding first maximum probability from the A side; the B fragment identification information is composed of the atomic identifier and the corresponding second maximum probability from the B side; and the binding site identification information is composed of the A and B fragment identification information and fed back to the current user.

[0058] A second aspect of the present invention provides an apparatus for implementing the processing method for predicting the binding sites of molecular fragments as described in the first aspect above, the apparatus comprising: a model building module, a model training module, and a model application module;

[0059] The model building module is used to build a binding site prediction model; the binding site prediction model is used to perform binding site prediction processing on the first and second molecular fragments input to the model and output the corresponding binding probability prediction vector.

[0060] The model training module is used to construct a model training dataset, denoted as the corresponding first dataset; and to train the binding site prediction model based on the first dataset;

[0061] The model application module is used to receive two three-dimensional molecular fragment structures as corresponding fragment structures A and B from the user after the model training is completed; and to use the binding site prediction model to identify the binding site information of the fragment structures A and B and to feed back the identification information to the current user.

[0062] A third aspect of the present invention provides an electronic device, including: a memory, a processor, and a transceiver;

[0063] The processor is used to couple with the memory, read and execute instructions in the memory to implement the steps of the method described in the first aspect above;

[0064] The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

[0065] A fourth aspect of the present invention provides a computer-readable storage medium storing computer instructions that, when executed by a computer, cause the computer to perform the instructions described in the first aspect.

[0066] This invention provides a method, apparatus, electronic device, and computer-readable storage medium for predicting binding sites of molecular fragments. As described above, this invention constructs a binding site prediction model that performs binding site prediction processing on first and second molecular fragments input to the model and outputs corresponding binding probability prediction vectors. A first dataset is constructed through big data collection to train the binding site prediction model. After model training, the binding site prediction model is used to identify the binding site information of any two three-dimensional molecular fragment structures input by the user and the identified information is fed back to the current user. The end-to-end binding site prediction model provided by this invention improves prediction efficiency and, through big data training, enhances the model's generalization ability and its ability to discover unknown connectivity patterns. Attached Figure Description

[0067] Figure 1 This is a schematic diagram of a method for predicting the binding sites of molecular fragments according to Embodiment 1 of the present invention.

[0068] Figure 2 This is a block diagram of the binding site prediction model provided in Embodiment 1 of the present invention;

[0069] Figure 3 This is a module structure diagram of a processing device for predicting the binding sites of molecular fragments provided in Embodiment 2 of the present invention;

[0070] Figure 4 This is a schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention. Detailed Implementation

[0071] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0072] Embodiment 1 of the present invention provides a method for predicting the binding sites of molecular fragments, such as... Figure 1 The schematic diagram shows a method for predicting the binding sites of molecular fragments according to Embodiment 1 of the present invention. This method mainly includes the following steps:

[0073] Step 1: Construct a binding site prediction model.

[0074] Here, the binding site prediction model of this invention is used to perform binding site prediction processing on the first and second molecular fragments input to the model and output the corresponding binding probability prediction vector.

[0075] It should be noted that the first and second molecular fragments input to the model are each a three-dimensional molecular fragment structure; here, the first and second molecular fragments can be two small fragments with a small number of atoms, or they can be two synthetic building blocks with a large number of atoms that also contain multiple small fragments; the data structures of the first and second molecular fragments are shown below:

[0076] 1) The first molecular segment includes the first set of atoms and the first set of chemical bonds;

[0077] The first atom set includes multiple first atoms; the atomic properties of a first atom include at least a first atom identifier, a first atom type, a first atom coordinates, and a first atom chiral property set; the first atom identifier is the unique identifier of the current first atom; the first atom type is the chemical element type of the current first atom; the first atom coordinates are the three-dimensional spatial coordinates of the current first atom; the first atom chiral property set consists of multiple first chiral properties; each first chiral property consists of a corresponding property type and property value; if the first atom is a chiral center atom, then all property values ​​in the corresponding first atom chiral property set are not empty; if the first atom is not a chiral center atom, then all property values ​​in the corresponding first atom chiral property set are empty;

[0078] The first set of chemical bonds includes multiple first chemical bonds; the chemical bond properties of the first chemical bond include at least the first chemical bond identifier, the first chemical bond type, and the first bonding atom group; the first chemical bond identifier is a unique identifier for the current first chemical bond; the first chemical bond type includes multiple chemical bond types; the first bonding atom group is composed of the first atom identifiers of the two first atoms connected by the current first chemical bond;

[0079] 2) The second molecular segment includes a second set of atoms and a second set of chemical bonds;

[0080] The second atom set includes multiple second atoms; the atomic properties of a second atom include at least a second atom identifier, a second atom type, a second atom coordinates, and a second atom chiral property set; the second atom identifier is the unique identifier of the current second atom; the second atom type is the chemical element type of the current second atom; the second atom coordinates are the three-dimensional spatial coordinates of the current second atom; the second atom chiral property set consists of multiple second chiral properties; each second chiral property consists of a corresponding property type and property value; if the second atom is a chiral center atom, then all property values ​​in the corresponding second atom chiral property set are not empty; if the second atom is not a chiral center atom, then all property values ​​in the corresponding second atom chiral property set are empty;

[0081] The second chemical bond set includes multiple second chemical bonds; the chemical bond properties of the second chemical bond include at least the second chemical bond identifier, the second chemical bond type, and the second bonding atom group; the second chemical bond identifier is the unique identifier of the current second chemical bond; the second chemical bond type includes multiple chemical bond types; the second bonding atom group is composed of the second atom identifiers of the two second atoms connected by the current second chemical bond.

[0082] It should also be noted that the first and second model input terminals of the binding site prediction model are used to receive the first and second molecular fragments as model inputs, respectively, and the model output terminal is used to output the corresponding binding probability prediction vector; such as Figure 2The block diagram of the binding site prediction model provided in Embodiment 1 of the present invention is shown.

[0083] The model components of the site prediction model include an encoding initialization module, a graph neural coding network, and an MLP network.

[0084] The connection relationships of each model component are as follows: the first and second input terminals of the encoding initialization module are connected to the first and second model input terminals respectively, and the output terminal is connected to the input terminal of the graph neural coding network; the output terminal of the graph neural coding network is connected to the input terminal of the MLP network; and the output terminal of the MLP network is connected to the model output terminal.

[0085] The functions of each model component are shown below.

[0086] 1) Encoding initialization module:

[0087] The encoding initialization module of the site prediction model is used to construct the molecular graph G based on the first and second molecular fragments input by the model; and to perform state encoding initialization processing based on the molecular graph G to obtain the corresponding state encoding tensor H0, which is then sent to the graph neural encoding network.

[0088] Here, the molecular graph G in this embodiment of the invention includes a node set V and an edge set E; wherein, the node set V consists of N1+N2 nodes v i Composition; the edge set E consists of multiple edges e i,j Composition; 1 ≤ node index i ≤ N1 + N2, 1 ≤ node index j ≤ N1 + N2, j ≠ i, N1 and N2 are the total number of atoms in the first and second molecular segments, respectively; node v i Corresponding to either a first atom or a second atom; edge e i,j It corresponds to a first chemical bond or a chemical bond.

[0089] The state encoding tensor H0 of this embodiment includes N1+N2 node state vectors h 0,i Node state vector h 0,i With node v i One-to-one correspondence; node state vector h 0,i It consists of a set of atomic type features, atomic coordinate features, a set of chiral features, a set of associated chemical bond features, and a set of cross-segment distance features; among which:

[0090] a. The atomic type characteristic is the current node's state vector h. 0,i The one-hot encoding of the first or second atom type corresponding to the first or second atom;

[0091] b. The atomic coordinate feature is the current node's state vector h. 0,i The corresponding first or second atom coordinates;

[0092] c. The chiral feature set consists of multiple chiral attribute features; each chiral attribute feature is the current node's state vector h. 0,i The feature encoding vector of a first or second chiral attribute in the set of chiral attributes of the first or second atom corresponding to the first or second atom;

[0093] d. The set of associated chemical bond features consists of one or more associated chemical bond features; associated chemical bond features include chemical bond type features and chemical bond direction features; each associated chemical bond feature is related to the current node's state vector h. 0,i The first or second atom corresponds to a first or second chemical bond; the chemical bond type feature is the unique thermal encoding of the first or second chemical bond type of the current first or second chemical bond; the chemical bond direction feature is a three-dimensional displacement vector from the current first or second atom to another first or second atom indicated in the first or second bonding atom group of the current first or second chemical bond.

[0094] e. The cross-segment distance feature set consists of multiple cross-segment distance features; if the current node state vector h 0,i For each first atom, the distance feature across segments is the square of the straight-line distance between the current first atom and a second atom in the second segment; if the current node state vector h 0,i For each second atom, each intermolecular distance feature is the square of the straight-line distance between the currently corresponding second atom and a first atom in the first molecule segment.

[0095] 2) Graph neural coding network:

[0096] The graph neural coding network that combines the site prediction model is composed of M layers of graph neural network (GNN) connected sequentially; where the total number of coding layers M is a preset positive integer with a value greater than 2.

[0097] The m-th GNN layer is used to encode the state tensor H input to this layer. m-1 Perform state transition processing to obtain the corresponding state encoding tensor H m And when the current coding layer index m is less than M, the state coding tensor H is... m Send to the (m+1)th GNN layer, and when the current coding layer index m equals M, convert the state coding tensor H. m＝M Send to the MLP network;

[0098] Here, 1 ≤ coding layer index m ≤ M; state coding tensor H m Includes N1+N2 node state vectors h m,iNode state vector h m,i With node v i One-to-one correspondence.

[0099] Node state vector h m,i The reasoning process is as follows:

[0100]

[0101] Among them, feature d i,j The setting rule is: if node v i v j If the two corresponding atoms belong to the same molecular segment, then the corresponding characteristic d i,j Fixed to 0; if node v i v j If the two corresponding atoms belong to two different molecular segments, then the corresponding characteristic d i,j Fixed as the corresponding cross-segment distance feature. f m Let θ be the state transition function of the m-th GNN layer. m f is the m-th state transition function m The model parameters of the corresponding m-th GNN layer.

[0102] 3) MLP network:

[0103] An MLP network combining a site prediction model is used to encode the state tensor H based on the output of the Mth GNN layer. M The corresponding combination probability prediction vector is obtained by performing combination probability prediction.

[0104] Here, in this embodiment of the invention, the combined probability prediction vector is predicted by N1+N2 nodes, resulting in probability p. i Composition; Node prediction probability p i With node v i One-to-one correspondence;

[0105] The reasoning process combining probability prediction vectors is: P = f MLP (H M ;θ MLP );

[0106] Where P is the combined probability prediction vector, f MLP For the network mapping function of the MLP network, θ MLP These are the model parameters for the MLP network.

[0107] Step 2: Construct a model training dataset, denoted as the first dataset; and train the binding site prediction model based on the first dataset.

[0108] Specifically, this includes: Step 21, constructing the model training dataset, denoted as the corresponding first dataset;

[0109] The first dataset includes multiple first data records; each first data record includes a first training molecular fragment, a second training molecular fragment, and a first label vector; the first and second training molecular fragments are each a three-dimensional molecular fragment structure; the first label vector includes multiple first node labels; the total number of first node labels is the sum of the total number of atoms in the first and second training molecular fragments; the label value of the first node label is 1 or 0; only two first node labels in the first label vector have a label value of 1, and the rest are 0, and the two atoms corresponding to the two first node labels with a label value of 1 are not in the same training molecular fragment;

[0110] The current step 21 specifically includes:

[0111] Step 211: Collect data on the three-dimensional molecular structures of various compound molecules through multiple publicly available data channels to obtain multiple corresponding first molecular structures;

[0112] Among them, various public data channels include at least various public datasets that publish molecular three-dimensional structures, journals / magazines / scientific and technological literature / papers;

[0113] Step 212: Each first molecular structure is taken as the corresponding current molecular structure; molecular fragment sampling is performed on the current molecular structure to obtain multiple corresponding sampling fragments; all sampling fragments of the current molecular structure are traversed once; during this traversal, the currently traversed sampling fragment is taken as the corresponding current fragment; other sampling fragments in the current molecular structure that are connected to the current fragment in a single-atom connection mode are extracted to form a corresponding associated fragment set; the current fragment is taken as a corresponding first training molecular fragment, and any sampling fragment in the current associated fragment set is taken as a corresponding second training molecular fragment, and the first training molecular fragment corresponding to the current fragment and each second training molecular fragment are combined to form a corresponding first fragment group; at the end of this traversal, all the obtained first fragment groups are deduplicated.

[0114] Step 213: Perform a traversal of all first fragment groups; during this traversal, the currently traversed first fragment group is taken as the corresponding current fragment group; calculate the total number of atoms in the first and second training molecular fragments of the current fragment group to obtain the corresponding current atomic sum; set a first label vector with a vector length matching the current atomic sum; initialize the label values ​​of all first node labels in the current first label vector to 0; after the vector initialization is completed, reset the label values ​​of the two first node labels corresponding to the two fragment connection atoms of the first and second training molecular fragments in the current fragment group to 1; and form a corresponding first data record from the first training molecular fragment, the second training molecular fragment, and the first label vector corresponding to the current fragment group; and at the end of this traversal, form the corresponding first dataset from all the obtained first data records.

[0115] Step 22, and train the binding site prediction model based on the first dataset;

[0116] Specifically, this includes: Step 221, dividing the first dataset into two sub-datasets based on a preset first segmentation ratio, denoted as the corresponding first training set and first evaluation set;

[0117] Wherein, the first segmentation ratio is a pre-set ratio parameter, such as 8:2; both the first training set and the first evaluation set consist of multiple first data records; the ratio of the total number of records in the first training set and the first evaluation set satisfies the first segmentation ratio;

[0118] Step 222: Extract the first data record of the first training set as the corresponding current training record;

[0119] Step 223: Input the first and second training molecular fragments of the current training record as the corresponding first and second molecular fragments into the binding site prediction model for binding site prediction processing, and use the binding probability prediction vector output by the model as the corresponding first prediction vector.

[0120] Step 224: Input the first predicted vector and the first label vector of the current training record into the preset first model loss function to calculate the corresponding first loss value;

[0121] The loss function of the first model includes at least the binary cross-entropy loss function;

[0122] Step 225: Identify whether the first loss value meets the preset first loss value range; if the first loss value meets the first loss value range, proceed to step 226; if the first loss value does not meet the first loss value range, based on the preset first model optimizer, perform a round of modulation on the model parameters of the graph neural coding network and the MLP network of the binding site prediction model in the direction of minimizing the first model loss function, and return to step 223 when the modulation ends.

[0123] Here, the first loss value range is a pre-set numerical range; the first model optimizer includes at least the SGD optimizer and the Adam optimizer;

[0124] Step 226: Identify whether the current training record is the last first data record of the first training set; if yes, proceed to step 227; if no, extract the next first data record of the first training set as the new current training record and return to step 223.

[0125] Step 227: Perform a round of traversal on all first data records in the first evaluation set; during this round of traversal, take the currently traversed first data record as the corresponding current evaluation record; input the first and second training molecular fragments of the current evaluation record as the corresponding first and second molecular fragments into the binding site prediction model for binding site prediction processing, and take the binding probability prediction vector output by the model as the corresponding second prediction vector; and form a corresponding prediction-label data pair by the second prediction vector and the first label vector of the current training record; and at the end of this round of traversal, calculate the accuracy, precision, recall and F1 score based on all the obtained prediction-label data pairs to obtain the corresponding first accuracy, first precision, first recall and first F1 score.

[0126] Step 228: Identify the first accuracy, first precision, first recall, and first F1 score; if the first accuracy does not meet the preset first accuracy range, or the first precision does not meet the preset first precision range, or the first recall does not meet the preset first recall range, or the first F1 score does not meet the preset first F1 score range, then return to step 222 to continue training; if the first accuracy meets the first accuracy range, and the first precision meets the first precision range, and the first recall meets the first recall range, and the first F1 score meets the first F1 score range, then stop training and confirm that the model training has ended.

[0127] Here, the first accuracy range, the first precision range, the first recall range, and the first F1 score range are four pre-set numerical ranges.

[0128] Step 3: After the model training is completed, the system receives two three-dimensional molecular fragment structures input by the user as corresponding fragment structures A and B; and uses the binding site prediction model to identify the binding site information of fragment structures A and B and feeds back the identification information to the current user.

[0129] Specifically, this includes: Step 31, after the model training is completed, receiving two three-dimensional molecular fragment structures input by the user as the corresponding fragment structures A and B;

[0130] Step 32: The binding site prediction model is used to identify the binding site information of fragment structures A and B, and the identification information is fed back to the current user.

[0131] Specifically, this includes: Step 321, inputting fragment structures A and B as the corresponding first and second molecular fragments into the binding site prediction model for binding site prediction processing, and using the binding probability prediction vector output by the model as the corresponding AB prediction vector;

[0132] Step 322: Calculate the predicted probabilities p of all nodes corresponding to all atoms in the AB prediction vector that are part of the fragment structure A. i Extract them to form the corresponding first probability set; and then predict the probability p of all nodes corresponding to all atoms of fragment structure B in the AB prediction vector. i Extract them to form the corresponding second probability set;

[0133] Step 323: Predict the probability p of the node with the highest probability value in the first probability set. i Let p be the corresponding first maximum probability; and predict the probability p of the node with the largest probability value in the second probability set. i This is denoted as the second maximum probability; and the atom identifier in fragment structure A corresponding to the first maximum probability is taken as the corresponding fragment A-side binding atom identifier; and the atom identifier in fragment structure B corresponding to the second maximum probability is taken as the corresponding fragment B-side binding atom identifier;

[0134] Step 324: The A fragment identification information is composed of the atomic identifier and the corresponding first maximum probability from the A side; the B fragment identification information is composed of the atomic identifier and the corresponding second maximum probability from the B side; and the binding site identification information is composed of the A and B fragment identification information and fed back to the current user.

[0135] Figure 3 This is a module structure diagram of a processing device for predicting the binding sites of molecular fragments according to Embodiment 2 of the present invention. This device can be a terminal device or server implementing the aforementioned method embodiments, or it can be a device that enables the aforementioned terminal device or server to implement the aforementioned method embodiments. For example, the device can be a device or chip system of the aforementioned terminal device or server. Figure 3 As shown, the device includes: a model building module 201, a model training module 202, and a model application module 203.

[0136] The model building module 201 is used to build a binding site prediction model; the binding site prediction model is used to perform binding site prediction processing on the first and second molecular fragments input to the model and output the corresponding binding probability prediction vector.

[0137] The model training module 202 is used to construct the model training dataset, denoted as the corresponding first dataset; and to train the binding site prediction model based on the first dataset.

[0138] The model application module 203 is used to receive two three-dimensional molecular fragment structures input by the user as corresponding fragment structures A and B after the model training is completed; and to use the binding site prediction model to identify the binding site information of fragment structures A and B and to feed back the identification information to the current user.

[0139] The present invention provides a processing device for predicting the binding sites of molecular fragments, which can execute the method steps in the above method embodiments. Its implementation principle and technical effect are similar, and will not be described again here.

[0140] It should be noted that the division of the various modules in the above device is merely a logical functional division. In actual implementation, they can be fully or partially integrated into a single physical entity, or they can be physically separated. Furthermore, these modules can be implemented entirely in software via processing elements; they can be fully implemented in hardware; or some modules can be implemented by processing elements calling software, while others are implemented in hardware. For example, the model building module can be a separate processing element, or it can be integrated into a chip in the above device. Alternatively, it can be stored as program code in the memory of the above device, and its functions can be called and executed by a processing element. The implementation of other modules is similar. Moreover, these modules can be fully or partially integrated together, or they can be implemented independently. The processing element described here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each of the above modules can be completed through integrated logic circuits in the hardware of the processor element or through software instructions.

[0141] For example, these modules can be one or more integrated circuits configured to implement the above methods, such as one or more Application Specific Integrated Circuits (ASICs), one or more Digital Signal Processors (DSPs), or one or more Field Programmable Gate Arrays (FPGAs). As another example, when a module is implemented using processing element scheduler code, the processing element can be a general-purpose processor, such as a Central Processing Unit (CPU) or other processor capable of calling program code. Furthermore, these modules can be integrated together as a System-on-a-Chip (SOC).

[0142] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. This computer program product includes one or more computer instructions. When these computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the foregoing method embodiments are generated. The computer described above can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The aforementioned computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the aforementioned computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, Bluetooth, microwave, etc.) means. The aforementioned computer-readable storage medium can be any available medium that a computer can access, or a data storage device such as a server or data center that integrates one or more available media. The aforementioned available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state disks (SSDs)).

[0143] Figure 4 This is a schematic diagram of an electronic device provided in Embodiment 3 of the present invention. This electronic device can be a terminal device or server implementing the methods of the aforementioned embodiments, or it can be a terminal device or server connected to the aforementioned terminal device or server implementing the methods of the aforementioned embodiments. Figure 4As shown, the electronic device may include: a processor 301 (e.g., CPU), a memory 302, and a transceiver 303; the transceiver 303 is coupled to the processor 301, and the processor 301 controls the transmission and reception operations of the transceiver 303. The memory 302 may store various instructions for performing various processing functions and implementing the processing steps described in the foregoing embodiments. Preferably, the electronic device involved in the embodiments of the present invention further includes: a power supply 304, a system bus 305, and a communication port 306. The system bus 305 is used to realize communication connections between components. The communication port 306 is used for communication between the electronic device and other peripherals.

[0144] exist Figure 4 The system bus 305 mentioned can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This system bus can be divided into address bus, data bus, control bus, etc. For ease of representation, it is represented by only one thick line in the figure, but this does not indicate that there is only one bus or one type of bus. The communication interface is used to enable communication between the database access device and other devices (e.g., clients, read-write libraries, and read-only libraries). Memory may include Random Access Memory (RAM) and may also include non-volatile memory, such as at least one disk storage device.

[0145] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), graphics processing units (GPUs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0146] It should be noted that the embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when run on a computer, cause the computer to perform the methods and processes provided in the above embodiments.

[0147] This invention provides a method, apparatus, electronic device, and computer-readable storage medium for predicting binding sites of molecular fragments. As described above, this invention constructs a binding site prediction model that performs binding site prediction processing on first and second molecular fragments input to the model and outputs corresponding binding probability prediction vectors. A first dataset is constructed through big data collection to train the binding site prediction model. After model training, the binding site prediction model is used to identify the binding site information of any two three-dimensional molecular fragment structures input by the user and the identified information is fed back to the current user. The end-to-end binding site prediction model provided by this invention improves prediction efficiency and, through big data training, enhances the model's generalization ability and its ability to discover unknown connectivity patterns.

[0148] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented in hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0149] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for predicting the binding sites of molecular fragments, characterized in that, The method includes: A binding site prediction model is constructed; the binding site prediction model is used to perform binding site prediction processing on the first and second molecular fragments input to the model and output the corresponding binding probability prediction vector. The model training dataset is denoted as the first dataset; and the binding site prediction model is trained based on the first dataset. After the model training is completed, it receives two three-dimensional molecular fragment structures input by the user as corresponding fragment structures A and B; and uses the binding site prediction model to identify the binding site information of fragment structures A and B and feeds back the identification information to the current user. The binding site prediction model includes an encoding initialization module, a graph neural coding network, and an MLP network. The encoding initialization module is used to construct a molecular graph G based on the first and second molecular fragments input from the model; and to perform state encoding initialization processing based on the molecular graph G to obtain a corresponding state encoding tensor H0, which is then sent to the graph neural coding network. The graph neural network is composed of M GNN layers connected sequentially, where the total number of coding layers M is a preset positive integer greater than 2; the m-th GNN layer is used to encode the state tensor H input to this layer. m-1 Perform state transition processing to obtain the corresponding state encoding tensor H m And when the current coding layer index m is less than M, the state coding tensor H is... m Send to the (m+1)th GNN layer, and when the current coding layer index m equals M, send the state coding tensor H. m＝M Send to the MLP network; 1 ≤ coding layer index m ≤ M; The MLP network is used to encode the state tensor H based on the output of the Mth layer of the GNN layer. M The corresponding combination probability prediction vector is obtained by performing combination probability prediction. The molecular graph G includes a node set V and an edge set E; The node set V consists of N1+N2 nodes v i Composition; the edge set E consists of multiple edges e i,j Composition; 1 ≤ node index i ≤ N1 + N2, 1 ≤ node index j ≤ N1 + N2, j ≠ i, N1 and N2 are the total number of atoms in the first and second molecular segments, respectively; the node v i Corresponding to one of the first atoms or one of the second atoms; the edge e i,j Corresponding to one of the first chemical bonds or one of the chemical bonds; The state encoding tensor H0 includes N1+N2 node state vectors h 0,i The node state vector h 0,i With the node v i One-to-one correspondence; the node state vector h 0,i It consists of a set of atomic type features, atomic coordinate features, chiral feature sets, associated chemical bond feature sets, and cross-segment distance feature sets; The cross-segment distance feature set consists of multiple cross-segment distance features; if the current node state vector h 0,i For each of the first atoms, each cross-segment distance feature is the square of the straight-line distance between the currently corresponding first atom and a second atom in the second molecular segment; if the current node state vector h 0,i For each second atom, each cross-segment distance feature is the square of the straight-line distance between the currently corresponding second atom and a first atom in the first molecular segment.

2. The processing method for predicting the binding sites of molecular fragments according to claim 1, characterized in that, The first and second molecular fragments are each a three-dimensional molecular fragment structure; The first molecular fragment includes a first set of atoms and a first set of chemical bonds; The first set of atoms includes multiple first atoms; the atomic attributes of the first atom include at least a first atom identifier, a first atom type, a first atom coordinates, and a first atom chiral attribute set; the first atom identifier is the unique identifier of the current first atom; the first atom type is the chemical element type of the current first atom; the first atom coordinates are the three-dimensional spatial coordinates of the current first atom; the first atom chiral attribute set consists of multiple first chiral attributes; each first chiral attribute consists of a corresponding attribute type and attribute value; if the first atom is a chiral center atom, then all attribute values ​​in the corresponding first atom chiral attribute set are not empty; If the first atom is not a chiral center atom, then all attribute values ​​in the corresponding chiral attribute set of the first atom are empty; The first set of chemical bonds includes multiple first chemical bonds; the chemical bond attributes of the first chemical bond include at least a first chemical bond identifier, a first chemical bond type, and a first bonding atom group; the first chemical bond identifier is a unique identifier for the current first chemical bond; the first chemical bond type includes multiple chemical bond types; the first bonding atom group is composed of the first atom identifiers of the two first atoms connected by the current first chemical bond; The second molecular fragment includes a second set of atoms and a second set of chemical bonds; The second atom set includes multiple second atoms; the atomic attributes of the second atom include at least a second atom identifier, a second atom type, second atom coordinates, and a second atom chiral attribute set; the second atom identifier is the unique identifier of the current second atom; the second atom type is the chemical element type of the current second atom; the second atom coordinates are the three-dimensional spatial coordinates of the current second atom; the second atom chiral attribute set consists of multiple second chiral attributes; each second chiral attribute consists of a corresponding attribute type and attribute value; if the second atom is a chiral center atom, then all attribute values ​​in the corresponding second atom chiral attribute set are not empty; If the second atom is not a chiral center atom, then all attribute values ​​in the corresponding chiral attribute set of the second atom are empty; The second set of chemical bonds includes multiple second chemical bonds; the chemical bond attributes of the second chemical bond include at least a second chemical bond identifier, a second chemical bond type, and a second bonding atom group; the second chemical bond identifier is a unique identifier for the current second chemical bond; the second chemical bond type includes multiple chemical bond types; the second bonding atom group is composed of the second atom identifiers of the two second atoms connected by the current second chemical bond.

3. The processing method for predicting the binding sites of molecular fragments according to claim 2, characterized in that, The first and second model input terminals of the binding site prediction model are used to receive the first and second molecular fragments input into the model, respectively, and the model output terminal is used to output the corresponding binding probability prediction vector. The first and second input terminals of the encoding initialization module are connected to the first and second model input terminals, respectively, and the output terminal is connected to the input terminal of the graph neural network; the output terminal of the graph neural network is connected to the input terminal of the MLP network; the output terminal of the MLP network is connected to the model output terminal. The atom type feature is the current node state vector h. 0,i The one-hot encoding of the first or second atom type corresponding to the first or second atom; The atomic coordinate feature is the current node state vector h. 0,i The corresponding atomic coordinates of the first or second atom; The chiral feature set consists of multiple chiral attribute features; each chiral attribute feature is the current node state vector h. 0,i The feature encoding vector of one of the first or second chiral attributes in the set of chiral attributes of the first or second atom corresponding to the first or second atom; The set of associated chemical bond features consists of one or more associated chemical bond features; the associated chemical bond features include chemical bond type features and chemical bond orientation features; each associated chemical bond feature is associated with the current node state vector h. 0,i The first or second atom corresponds to a first or second chemical bond; the chemical bond type feature is the unique thermal encoding of the first or second chemical bond type of the currently corresponding first or second chemical bond; The chemical bond orientation feature is a three-dimensional displacement vector starting from the currently corresponding first or second atom and ending at another first or second atom indicated in the first or second bonding atom group of the currently corresponding first or second chemical bond. The state encoding tensor H m Includes N1+N2 node state vectors h m,i The node state vector h m,i With the node v i One-to-one correspondence; Feature d i,j The setting rule is: if node v i v j If the two corresponding atoms belong to the same molecular segment, then the corresponding characteristic d i,j Fixed to 0; if node v i v j If the two corresponding atoms belong to two different molecular segments, then the corresponding characteristic d i,j Fixed to the corresponding cross-segment distance feature; f m Let θ be the state transition function of the m-th layer of the GNN. m f is the m-th state transition function m The corresponding model parameters of the m-th GNN layer; The combined probability prediction vector is predicted by N1+N2 nodes, resulting in probability p. i Composition; the node prediction probability p i With the node v i One-to-one correspondence; P=f MLP (H M ;θ MLP ); P is the combined probability prediction vector, f MLP Let θ be the network mapping function of the MLP network. MLP These are the model parameters of the MLP network.

4. The processing method for predicting the binding sites of molecular fragments according to claim 3, The first dataset includes multiple first data records; each first data record includes a first training molecular fragment, a second training molecular fragment, and a first label vector. The first and second training molecular fragments are each a three-dimensional molecular fragment structure; The first label vector includes multiple first node labels; the total number of first node labels is the sum of the total number of atoms in the first and second training molecular segments; the label value of the first node label is 1 or 0; only two first node labels in the first label vector have a label value of 1, and the rest are 0, and the two atoms corresponding to the two first node labels with a label value of 1 are not in the same training molecular segment.

5. The processing method for predicting the binding sites of molecular fragments according to claim 4, wherein the model training dataset is denoted as the corresponding first dataset, specifically including; Step 51: Collect data on the three-dimensional molecular structures of various compound molecules through multiple publicly available data channels to obtain multiple corresponding first molecular structures; in, The aforementioned multiple types of publicly available data channels include various publicly available datasets that publish the three-dimensional structure of molecules; Step 52: Each of the first molecular structures is taken as the corresponding current molecular structure; molecular fragment sampling is performed on the current molecular structure to obtain multiple corresponding sampling fragments; a round of traversal is performed on all the sampling fragments of the current molecular structure; during this round of traversal, the currently traversed sampling fragment is taken as the corresponding current fragment; other sampling fragments in the current molecular structure that are connected to the current fragment in a single-atom connection mode are extracted to form a corresponding associated fragment set; the current fragment is taken as a corresponding first training molecular fragment, and any sampling fragment in the current associated fragment set is taken as a corresponding second training molecular fragment, and the first training molecular fragment corresponding to the current fragment and each of the second training molecular fragments are combined to form a corresponding first fragment group; at the end of this round of traversal, all the obtained first fragment groups are deduplicated. Step 53: Perform one round of traversal on all the first segment groups; and during this round of traversal, take the first segment group that is currently being traversed as the corresponding current segment group; The total number of atoms in the first and second training molecular fragments of the current fragment group is calculated to obtain the corresponding current total number of atoms; And set a first label vector with a length matching the current atomic sum; and initialize the label values ​​of all first node labels in the current first label vector to 0; After the vector initialization is completed, the label values ​​of the two first node labels corresponding to the two fragment connection atoms of the first and second training molecular fragments in the current fragment group are reset to 1 in the first label vector; and the first training molecular fragment, the second training molecular fragment and the first label vector corresponding to the current fragment group are combined to form a corresponding first data record; At the end of this round of traversal, the first dataset is composed of all the first data records obtained.

6. The processing method for predicting the binding sites of molecular fragments according to claim 5, wherein the multiple types of publicly available data channels include scientific and technological literature; the scientific and technological literature includes journals, magazines, and papers.

7. The processing method for predicting binding sites of molecular fragments according to claim 4, wherein training the binding site prediction model based on the first dataset specifically includes: Step 61: Based on a preset first segmentation ratio, the first dataset is divided into two sub-datasets, denoted as the first training set and the first evaluation set. in, Both the first training set and the first evaluation set consist of multiple first data records; the ratio of the total number of records in the first training set and the first evaluation set satisfies the first segmentation ratio; Step 62: Extract the first data record of the first training set as the corresponding current training record; Step 63: Input the first and second training molecular fragments of the current training record as the corresponding first and second molecular fragments into the binding site prediction model for binding site prediction processing, and use the binding probability prediction vector output by the model as the corresponding first prediction vector. Step 64: Input the first predicted vector and the first label vector of the current training record into a preset first model loss function to calculate the corresponding first loss value; Wherein, the loss function of the first model includes at least the binary cross-entropy loss function; Step 65: Identify whether the first loss value meets the preset first loss value range; if the first loss value meets the first loss value range, proceed to step 66; if the first loss value does not meet the first loss value range, modulate the model parameters of the graph neural coding network and the MLP network of the binding site prediction model in one round based on the preset first model optimizer in the direction of minimizing the first model loss function, and return to step 63 when the modulation ends. Step 66: Identify whether the current training record is the last first data record in the first training set; if yes, proceed to step 67; if no, extract the next first data record in the first training set as the new current training record and return to step 63. Step 67: Perform a round of traversal on all the first data records in the first evaluation set; during this round of traversal, take the currently traversed first data record as the corresponding current evaluation record; input the first and second training molecular fragments of the current evaluation record as the corresponding first and second molecular fragments into the binding site prediction model for binding site prediction processing, and take the binding probability prediction vector output by the model as the corresponding second prediction vector; and form a corresponding prediction-label data pair by the second prediction vector and the first label vector of the current training record; and at the end of this round of traversal, calculate the accuracy, precision, recall and F1 score based on all the obtained prediction-label data pairs to obtain the corresponding first accuracy, first precision, first recall and first F1 score; Step 68: Identify the first accuracy, first precision, first recall, and first F1 score; if the first accuracy does not meet the preset first accuracy range, or the first precision does not meet the preset first precision range, or the first recall does not meet the preset first recall range, or the first F1 score does not meet the preset first F1 score range, then return to step 62 to continue training; if the first accuracy meets the first accuracy range, and the first precision meets the first precision range, and the first recall meets the first recall range, and the first F1 score meets the first F1 score range, then stop training and confirm that the model training has ended.

8. The processing method for predicting the binding sites of molecular fragments according to claim 3, wherein the step of using the binding site prediction model to identify the binding site information of fragment structures A and B and feeding back the identification information to the current user specifically includes: Step 71: Input the fragment structures A and B as the corresponding first and second molecular fragments into the binding site prediction model for binding site prediction processing, and use the binding probability prediction vector output by the model as the corresponding AB prediction vector. Step 72: Calculate the predicted probabilities p of all nodes in the AB prediction vector corresponding to all atoms of the fragment structure A. i Extract them to form the corresponding first probability set; and then extract all the node prediction probabilities p in the AB prediction vector corresponding to all atoms of the fragment structure B. i Extract them to form the corresponding second probability set; Step 73, predict the probability p of the node with the largest probability value in the first probability set. i Let p be the corresponding first maximum probability; and let the node with the largest probability value in the second probability set be predicted as p. i The corresponding second maximum probability is recorded; and the atom identifier in the fragment structure A corresponding to the first maximum probability is used as the corresponding fragment A-side binding atom identifier; and the atom identifier in the fragment structure B corresponding to the second maximum probability is used as the corresponding fragment B-side binding atom identifier; Step 74: The A fragment identification information is composed of the atomic identifier and the corresponding first maximum probability from the A side; the B fragment identification information is composed of the atomic identifier and the corresponding second maximum probability from the B side; and the binding site identification information is composed of the A and B fragment identification information and fed back to the current user.

9. An apparatus for performing the processing method for predicting the binding sites of molecular fragments according to any one of claims 1-8, characterized in that, The device includes: a model building module, a model training module, and a model application module; The model building module is used to build a binding site prediction model; the binding site prediction model is used to perform binding site prediction processing on the first and second molecular fragments input to the model and output the corresponding binding probability prediction vector. The model training module is used to construct a model training dataset, denoted as the corresponding first dataset; and to train the binding site prediction model based on the first dataset; The model application module is used to receive two three-dimensional molecular fragment structures as corresponding fragment structures A and B from the user after the model training is completed; and to use the binding site prediction model to identify the binding site information of the fragment structures A and B and to feed back the identification information to the current user.

10. An electronic device, characterized in that, include: Memory, processor, and transceiver; The processor is configured to be coupled to the memory, read and execute instructions in the memory to implement the method according to any one of claims 1-8; The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a computer, cause the computer to perform the method described in any one of claims 1-8.

Images